I don’t see anything in your post that indicates any reason to track what posts a person has read. That should not be tracked at all. Reading posts should be completely anonymous.
I don’t see why voting necessarily has to track who casts the votes. But, because untracked voting can be abused so easily, I can understand deciding to retain the info for let’s say 24 hours. Hopefully that is also enough to handle those propagation issues.
Really, imho, server instances shouldn’t have a web interface at all, just an API. Web apps would make API calls to the server and reformat the response for use by the browser. The API call to read a post should not require any identifying info or require the user to be logged in. Read tracking and subscriptions should be handled by the client, and in the case of a public client (web app shared by many users), the private user info should be encrypted in case of a server breakin or seizure. The encryption key would be based on the user password and transformed to a browser cookie when the user logs in, so it is never stored by the web app. With most people using mobile clients these days, alternatively, the info can be kept completely on the client device and maintained by the mobile app.